OpenAI breaks new ground as the artificial intelligence giant revealed its first text-to-video model called Sora, capable of creating strikingly realistic content.
We've been wondering when the company will finally launch its own video engine, as many of its rivals, from Stability AI to Google, have beaten it to the punch. Perhaps OpenAI wanted to get things right before a proper launch. At this rate, the quality of their products could eclipse that of their contemporaries. According to the official website, Sora can generate “realistic and imaginative scenes” from a single text message; Just like other text-to-video AI models. The difference with this engine is the technology behind it.
Realistic content
Open AI claims its artificial intelligence can understand how people and objects “exist in the physical world.” This gives Sora the ability to create scenes with multiple people, different types of movement, facial expressions, textures, and objects with a high amount of detail. The generated videos lack the plastic look or nightmarish shapes seen in other AI content, for the most part, but more on that later.
Sora is also multi-modular. Users will reportedly be able to upload a still image to serve as the basis for a video. The content within the image will be animated by paying close attention to small details. You can even take a pre-existing video “and enlarge it or fill in the missing frames.”
Notice: A litter of golden retriever puppies playing in the snow. Their heads emerge from the snow, covered in snow. pic.twitter.com/G1qhJRV9tgFebruary 15, 2024
You can find sample clips on the OpenAI website and on X (the platform formerly known as Twitter). One of our favorites features a group of puppies playing in the snow. If you look closely, you can see that their fur and the snow on their snouts have a surprisingly realistic quality. Another great video shows a dove with a Victoria crown moving like a real bird.
A work in progress
As impressive as these two videos may be, Sora is not perfect. OpenAI admits that its “model has weaknesses.” You may have difficulty simulating the physics of an object, confusing left with right, and misunderstanding “cases of cause and effect.” You can have an AI character bite into a cookie, but the cookie lacks a bite mark.
He also makes a lot of strange mistakes. One of the funniest mishaps. It involves a group of archaeologists unearthing a large piece of paper which is then transformed into a chair before ending up as a crumpled piece of plastic. The AI also seems to have trouble with words. “Otter” is misspelled as “Oter” and “Land Rover” is now “Danover.”
Even Sora's mistakes are fascinating pic.twitter.com/OvPSbaa0L9February 15, 2024
Going forward, the company will work with its “red teams,” which are a group of industry experts “to assess critical areas for damage or risks.” They want to make sure Sora doesn't create false information, hateful content, or bias. Additionally, OpenAI will implement a text classifier to reject messages that violate its policy. These include entries soliciting sexual content, violent videos, and portraits of celebrities, among other things.
It is unknown when Sora will be officially released. We're reaching out for release information. This story will be updated later. In the meantime, check out TechRadar's list of the best AI video editors for 2024.
you might also like
